The Foundation Model for Crypto
Blockchains like Ethereum have set the stage for a digital revolution by creating an infrastructural base layer for compute. This ever-evolving graph of nodes, accounts, and smart contracts is now uniquely positioned to become the native domain for decentralized AI.
But while blockchains provide the native rails to underpin this future, crypto still needs its own foundation model designed from first principles for decentralized AI to realize its full potential.
Pond is already creating this final piece of the puzzle—an onchain model layer supported by an ecosystem dedicated to the collective development and ownership of powerful crypto-native AI models.
At Archetype, we have been fascinated by first principles approaches to marrying blockchains and AI, and Pond’s crypto-native approach resonated with us from our first conversation.
Onchain AI’s Missing Puzzle Piece
The foundation model being developed by Pond uses a graph-based neural network (GNN) architecture specifically built to unlock a deep understanding of onchain data. Pond’s approach uniquely illuminates the intricate dependencies and interactions between nodes, accounts, and transactions to enable novel predictive capabilities that set the stage for vast new use cases.
Where conventional LLMs predict words based on historical inputs, Pond’s GNN processes and extracts graphical data to predict complex relationships from a blockchain’s graph structure and corresponding attributes.
As the agentic web continues to emerge atop blockchains—leveraging them for native orchestration and monetization—Pond’s intelligence layer bolsters this connection by enabling agents and applications to harness the symbiotic nature of these onchain relationships.
Pond’s model ecosystem ultimately seeks to unlock the agent economy with a dynamic intelligence layer that equips agents with unparalleled context and real-time insights into the foundational infrastructure on which they’ll natively run.
Pond is already using its model to expand the frontier of defi, from assessing market sentiment for price predictions to mitigating risk by analyzing unusual interaction frequencies, behaviors, and fund flows.
Likewise, Pond is designing a model for personalized recommendation mechanisms for onchain social networks with an architecture inspired by the same AI systems powering today’s web2 social media platforms.
A new wave of intelligent protocols and applications now able to tap into Pond’s crypto-native model layer are set to reimagine crypto’s user experience by unlocking transformative onchain capabilities on par with the best consumer applications in web2.
Building Crypto’s Foundational AI Model
We’re ecstatic to lead Pond’s $7.5M Seed round and accelerate their mission of building the foundational model for crypto.
Pond’s leadership is composed of elite technical talent including the former ML architect of WeChat, a former scientist at Amazon, and the former principal data engineer at Yuga Labs. This team collectively has over 40 research publications across AI journals and conferences like NeurIPS, ICML, PNAS, IEEE, Nature portfolios (Nature Human Behavior, Nature Communications), and Harvard Business Review.
Dedicated to democratizing access to its model and data infrastructure, Pond is empowering the entire crypto ecosystem to advance AI in groundbreaking ways. We’re thrilled to work alongside Dylan Zhang, Bill Shi, and the rest of the Pond team as they usher in this future.
To get involved with Pond’s model ecosystem, check out Frontier—an ongoing series of competitions to train the most powerful web3 models. Improve your skills and earn great bonuses at the same time!
